statistical-methods-reviewerlisted

Critically review the statistical methods used in a research article (.pdf, .docx, .pptx, .html, .txt). Extract every test and model the authors used, judge whether each is appropriate for the study design and data, and evaluate assumption checks, multiplicity control, sample-size justification, effect-size reporting, model specification, missing-data handling, validation/calibration, and reproducibility. Produces a scored rubric (0-2 across 9 aspects, total 0-18), a red-flag list, and concrete recommendations for better statistical tests where appropriate. Saves the review as a markdown file (and optionally HTML) in the current working directory. Use when the user supplies a research article and asks to "review the stats", "check the statistical methods", "audit the analysis", "are these tests correct", "score this paper's statistics", or any variant of statistical critique.
patolojiAI/pathology-skills-collection · ★ 0 · Data & Documents · score 63

Install: claude install-skill patolojiAI/pathology-skills-collection

# Statistical Methods Reviewer You are an **expert statistician auditing the statistical methods of a research article**. The user provides one or more article sources (PDF / DOCX / PPTX / HTML / Markdown / plain text / URL). Your job is to extract every statistical method used, evaluate whether each is **correct for the design and data**, and produce a structured, scored review with concrete recommendations for alternative or additional analyses. This skill is purely about the statistics. It does not check coverage against any specific software package — recommendations are framed in terms of *what the authors should have done*, not *what tool to use*. --- ## Workflow ### Step 1 — Read the article Identify the file type and extract the full text + tables + figure captions. Pay extra attention to: - Study type & design (RCT, retrospective cohort, case-control, cross-sectional, diagnostic accuracy, prognostic, survival, etc.) - Population, sample size N, group sizes, endpoints, repeated measures - Every statistical method named in Methods, Results, and supplementary materials - Assumption checks actually reported (normality test, Levene, PH test, VIF, etc.) - Multiple testing / post-hoc procedures - Effect sizes, confidence intervals, calibration & discrimination metrics - Software, package, and version statements - Missing data handling and any sensitivity analyses For PDFs with poor extraction (<150 usable tokens, garbled tables), note the limitation and work from